Image Processing Algorithm Research And Implimen-Tation Of Video Surveillance Based On FPGA

In recent years,with the rapid growth of China's security industry,the video monitoring system has been prevalent in human lives and production.As the related semiconductor technology development,the new processors and sensors have been constantly emerging,so the development of the video monitoring system has been moving towards the direction of high-resolution.The early monitoring system often has only the function of image capture,through transmitting the acquired video images to the back-end server for video detection and analysis.But now in order to improve the intelligent video monitoring system and reduce data transmission network with the overall system power consumption,it is of great significance that the analysis of the video image detection realized in front-end。Comply with the development trend of video monitoring system of security industry,In this paper,based on the actual product project,the research on key technology of high definition video image collection based on FPGA,at the same time based on the basic image processing algorithm for hardware implementation,complete the typical detection and analysis of application which based on the system of moving object detection.The main research work of this paper contains:(1)Image acquisition is based on FPGA and CCD framework.Firstly,the the digital signal processor chip is configured by three-wire serial to drive the CCD to collect signals,then the signal form the CCD is processed by sample,analog-digital conversion and so on in AD9928.After that the processed signal is transfered to FPGA,the received signal is processed by FPGA through the following transformations sequentially: LVDS decoding,image stitching cache,color filter and YUV color space conversion and is output finally.(2)Design the testbench of FPGA to simulate image video circuit,and design circuit of commonly used image processing algorithm,includes,color space conversion,template filtering,mathematical morphology dilation and erosion,integral image,local binary patternsand so on.(3)For the problem of moving object detection algorithm,which is that sensitive to light and noise,an algorithm for moving object detection combining pixel space with texture features is proposed in this paper and realized based on FPGA.The foreground is detected by the confidence factor of neighborhood pixels and the background is update through Center-symmetric Local Binary Pattern.The experiment shows that the algorithm in this paper with completely target extraction is robust to illumination whose result is superior to the traditional Mixed Gaussian Model.In this paper,through the research on these key technologies,the design of circuit module and debugging of interface circuit,so that the whole system can work normally with the high definition video image output,whose frame rate is sixteen and resolution is 8 million,at the same time the system can be effectively used to detect the moving target,meet the actual use demand.The system has fast processing speed and low power consumption.